Li Shuaiqi's Blog HomeCategoriesTagsArchivesAbout
Process篇 6 - 从 dumpsys meminfo 看进程的优先级
|AndroidFramework源码分析Process进程
字数总计: 1.8k|阅读时长: 9 分钟

0 前言

基于 android 7.1.1 源码，分析和总结 Process 相关知识！

1 dumpsys meminfo –oom

dumpsys meminfo 可以来看系统的内存使用情况，这里我们重点关注：Total PSS by OOM adjustment:

默认的 dumpsys meminfo 是出了可以看 Total PSS by OOM adjustment，还可以看 Total PSS by process 等等详细的信息，这里我们只关注第一种，这里有一个很方便的指令：

adb shell dumpsys meminfo --oom

下面是该指令的输出：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
Applications Memory Usage (in Kilobytes):
Uptime: 37282469 Realtime: 62579406

     ... ... ... ...

Total PSS by OOM adjustment:
    304,190K: Native
         38,170K: android.hardware.camera.provider@2.4-service (pid 683)
         17,815K: logd (pid 539)
         15,833K: vendor.oppo.hardware.biometrics.face@1.0-service (pid 1109)
         15,014K: surfaceflinger (pid 721)
          7,927K: android.hardware.audio@2.0-service (pid 681)
          7,318K: webview_zygote32 (pid 2635)
          7,105K: rild (pid 1083)
          6,857K: media.codec (pid 1082)
          ... ... ... ...
  
    119,314K: System
        119,314K: system (pid 1703)

    168,951K: Persistent
         94,755K: com.android.systemui (pid 2598)
         23,766K: com.android.phone (pid 2738)
         11,445K: .dataservices (pid 2706)
          7,254K: com.oppo.multimedia.dirac (pid 3493)
          ... ... ... ...

     95,474K: Foreground
         33,792K: com.coloros.safecenter:clear_filter (pid 2514)
         26,509K: com.oppo.ota:ui (pid 15668 / activities)
         13,879K: com.coloros.securitypermission (pid 3140)
         12,447K: com.oppo.ota (pid 15686)
          8,847K: com.nearme.romupdate (pid 2491)
          ... ... ... ...
          
    524,433K: Visible
        105,184K: com.tencent.mm:push (pid 7125)
        100,550K: com.tencent.mm:exdevice (pid 5237)
         62,929K: com.oppo.launcher (pid 3325 / activities)
         27,815K: android.process.contacts (pid 4085)
         24,425K: com.coloros.safecenter (pid 3708)
         23,279K: com.coloros.mcs (pid 4122)
         21,670K: android.process.acore (pid 4310)
         20,998K: com.coloros.recents (pid 3267 / activities)
         ... ... ... ...
          
     39,159K: Perceptible
         31,468K: com.sohu.inputmethod.sogouoem (pid 2584)
          7,691K: com.amap.android.location (pid 4132)
          
    103,667K: A Services
         72,943K: com.kuaikan.comic:monitorService (pid 4938)
         30,724K: com.kuaikan.comic:QS (pid 5207)
         
    174,752K: Previous
        147,972K: com.tencent.mm (pid 7318)
         26,780K: com.android.settings (pid 13815 / activities)
         
    489,783K: B Services
        370,782K: com.kuaikan.comic (pid 4869 / activities)
         28,757K: com.android.mms (pid 5120)
         27,198K: com.kuaikan.comic:QALSERVICE (pid 5063)
         15,458K: com.coloros.selfcheck (pid 15626 / activities)
         15,422K: com.tencent.mobileqq:MSF (pid 10558)
         14,523K: com.nearme.statistics.rom (pid 2556)
          9,211K: com.coloros.gallery3d (pid 16053)
          4,513K: com.coloros.activation (pid 5301)
          3,919K: com.coloros.usbselection (pid 16123)
          
     76,443K: Cached
         39,483K: com.coloros.aiservice (pid 7038)
         17,041K: android.process.media (pid 13010)
         10,643K: com.android.providers.downloads (pid 3251)
          4,979K: com.coloros.securepay (pid 9412)
          4,297K: com.android.defcontainer (pid 15323)

Total RAM: 3,831,972K (status normal)
 Free RAM: 1,734,943K (   76,527K cached pss + 1,319,364K cached kernel +    24,584K ion cached +   314,468K free)
 Used RAM: 2,459,876K (2,083,100K used pss +   376,776K kernel)
 Lost RAM:  -979,823K
     ZRAM:   616,976K physical used for 1,361,056K in swap (2,097,148K total swap)
   Tuning: 384 (large 512), oom   322,560K, restore limit   107,520K (high-end-gfx)

我们可以看到，这部分，系统是根据进程的 adj 将其分类，然后根据不同的分类，显示出了不同类别进程的内存使用情况！

这里我们关心的是 adb shell dumpsys meminfo --oom 是如何归类不同类型的进程的！

2 dumpApplicationMemoryUsage

adb shell dumpsys meminfo --oom 最终会调用 dumpApplicationMemoryUsage 函数，这里我们重点关注和 --oom 相关的逻辑：

这里我们先来看几个和该指令相关的常量;

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
static final int[] DUMP_MEM_OOM_ADJ = new int[] {
        ProcessList.NATIVE_ADJ,
        ProcessList.SYSTEM_ADJ, ProcessList.PERSISTENT_PROC_ADJ,
        ProcessList.PERSISTENT_SERVICE_ADJ, ProcessList.FOREGROUND_APP_ADJ,
        ProcessList.VISIBLE_APP_ADJ, ProcessList.PERCEPTIBLE_APP_ADJ,
        ProcessList.BACKUP_APP_ADJ, ProcessList.HEAVY_WEIGHT_APP_ADJ,
        ProcessList.SERVICE_ADJ, ProcessList.HOME_APP_ADJ,
        ProcessList.PREVIOUS_APP_ADJ, ProcessList.SERVICE_B_ADJ, ProcessList.CACHED_APP_MIN_ADJ
};

static final String[] DUMP_MEM_OOM_LABEL = new String[] {
        "Native",
        "System", "Persistent", "Persistent Service", "Foreground",
        "Visible", "Perceptible",
        "Heavy Weight", "Backup",
        "A Services", "Home",
        "Previous", "B Services", "Cached"
};

DUMP_MEM_OOM_ADJ 中封装的是 oom adj，DUMP_MEM_OOM_LABEL 中封装的是 lebal，这个看命令输出，很容易猜到是什么意思！

下面我们来重点分析方法！！

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
final void dumpApplicationMemoryUsage(FileDescriptor fd,
        PrintWriter pw, String prefix, String[] args, boolean brief, PrintWriter categoryPw) {
        
    ... ... ... ...
    
    // 解析指令，因为我们有 `--oom`，所以 oomOnly 为 true！
    else if ("--oom".equals(opt)) {
            oomOnly = true;
        } 
        ... ... ... ...
    }

    long uptime = SystemClock.uptimeMillis();
    long realtime = SystemClock.elapsedRealtime();
    final long[] tmpLong = new long[1];

    // 收集进程，返回的就是 mLruProcesses 的拷贝！
    ArrayList<ProcessRecord> procs = collectProcesses(pw, opti, packages, args);

    ... ... ... ...
    
    // 创建一些数组，用于封装信息，数组长度均是 DUMP_MEM_OOM_LABEL.length: 14！
    long oomPss[] = new long[DUMP_MEM_OOM_LABEL.length];
    long oomSwapPss[] = new long[DUMP_MEM_OOM_LABEL.length];
    ArrayList<MemItem>[] oomProcs = (ArrayList<MemItem>[])
            new ArrayList[DUMP_MEM_OOM_LABEL.length];

    long totalPss = 0;
    long totalSwapPss = 0;
    long cachedPss = 0;
    long cachedSwapPss = 0;
    boolean hasSwapPss = false;

    Debug.MemoryInfo mi = null;
    
    // 逆序遍历进程集合！
    for (int i = procs.size() - 1 ; i >= 0 ; i--) {
        // 获得进程对象！
        final ProcessRecord r = procs.get(i);
        final IApplicationThread thread;
        final int pid;
        final int oomAdj;
        final boolean hasActivities;
        synchronized (this) {
            thread = r.thread;
            pid = r.pid; // 收集 pid！
            oomAdj = r.getSetAdjWithServices(); // 获得该进程的 oomAdj！
            hasActivities = r.activities.size() > 0; // 该进程是否正在运行 activity！
        }

        if (thread != null) {
            ... ... ... ...

            if (!isCheckinRequest && mi != null) {
                totalPss += myTotalPss;
                totalSwapPss += myTotalSwapPss;
                MemItem pssItem = new MemItem(r.processName + " (pid " + pid +
                        (hasActivities ? " / activities)" : ")"), r.processName, myTotalPss,
                        myTotalSwapPss, pid, hasActivities);

               ... ... ... ...
                
                //【1】便利 oomPss 数组，匹配 oomAdj！
                for (int oomIndex = 0; oomIndex < oomPss.length; oomIndex++) {

                    //【1.1】当此时已经到了 oomPss 的最后一个元素，或者
                    // 该进程的 oomAdj 处于 [ DUMP_MEM_OOM_ADJ[oomIndex], DUMP_MEM_OOM_ADJ[oomIndex + 1] ) 之间！
                    // 这是我们匹配到了合适的 oomAdj，进入以下逻辑！
                    if (oomIndex == (oomPss.length - 1)
                            || (oomAdj >= DUMP_MEM_OOM_ADJ[oomIndex]
                                    && oomAdj < DUMP_MEM_OOM_ADJ[oomIndex + 1])) {
                        
                        // oomPss[oomIndex] 累加上 myTotalPss！
                        oomPss[oomIndex] += myTotalPss;
                        // oomSwapPss[oomIndex] 累加上 myTotalSwapPss！
                        oomSwapPss[oomIndex] += myTotalSwapPss;

                        // 创建一个 ArrayList 列表，用于封装进程的内存对象！
                        if (oomProcs[oomIndex] == null) {
                            oomProcs[oomIndex] = new ArrayList<MemItem>();
                        }

                        // 将 pssItem 添加爱到该 ArrayList 中去！
                        oomProcs[oomIndex].add(pssItem);
                        break;
                    }
                }
            }
        }
    }
    
    
    // 下面这段逻辑很重要，决定了 adb shell dumpsys meminfo --oom 是如何对进程划分的！
    if (!isCheckinRequest && procs.size() > 1 && !packages) {
        ... ... ... ...

        ArrayList<MemItem> oomMems = new ArrayList<MemItem>();

        // 遍历 oomPss 数组
        for (int j = 0; j < oomPss.length; j++) {
            if (oomPss[j] != 0) {
                // 获得 label ，因为 isCompact 为 false，所以为 UMP_MEM_OOM_LABEL[j]！
                String label = isCompact ? DUMP_MEM_OOM_COMPACT_LABEL[j]
                        : DUMP_MEM_OOM_LABEL[j];

                // 创建总的 MemItem 对象，用于封装每种类型下的进程的信息；
                // oomPss[j] 表示这个 label 下的总物理内存；
                // oomSwapPss[j] 表示该 label 下的总交换物理内存；
                // DUMP_MEM_OOM_ADJ[j] 表示该进程类别的起始 oom adj；
                MemItem item = new MemItem(label, label, oomPss[j], oomSwapPss[j],
                        DUMP_MEM_OOM_ADJ[j]);
                        
                // subitems 则用来表示属于该类别的所有进程信息！
                item.subitems = oomProcs[j];
                oomMems.add(item);
            }
        }

        ... ... ... ...
        
        if (!isCompact) {
            pw.println("Total PSS by OOM adjustment:");
        }
        
        // 显示出最终的信息！！
        dumpMemItems(pw, "  ", "oom", oomMems, false, isCompact, dumpSwapPss);
 
        ... ... ... ... 
    }
}

上面的逻辑是和 Total PSS by OOM adjustment 相关的！

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
ArrayList<ProcessRecord> collectProcesses(PrintWriter pw, int start, boolean allPkgs,
        String[] args) {
    ArrayList<ProcessRecord> procs;
    synchronized (this) {
        if (args != null && args.length > start
                && args[start].charAt(0) != '-') {
            procs = new ArrayList<ProcessRecord>();
            int pid = -1;

            try {
                pid = Integer.parseInt(args[start]);
            } catch (NumberFormatException e) {
            }

            for (int i=mLruProcesses.size()-1; i>=0; i--) {
                ProcessRecord proc = mLruProcesses.get(i);
                if (proc.pid == pid) { // 按照 pid 收集；
                    procs.add(proc);

                } else if (allPkgs && proc.pkgList != null
                        && proc.pkgList.containsKey(args[start])) { // 按照 packageName 收集；
                    procs.add(proc);

                } else if (proc.processName.equals(args[start])) { // 按照 processName 收集！
                    procs.add(proc);

                }
            }

            if (procs.size() <= 0) {
                return null;
            }
        } else {
            // 不指定 pid，processName，packageName 的话，就返回 mLruProcesses！
            procs = new ArrayList<ProcessRecord>(mLruProcesses);
        }
    }
    return procs;
}

其实，我们可以看出 adb shell dumpsys meminfo --oom 没有传入任何参数，所以返回的就是 mLruProcesses 的拷贝！

3 总结

分析了 adb shell dumpsys meminfo –oom 方法的执行流程，我们来总结下，进程的分类：

序号 进程 Label 起始 oom adj 值
1 Native ProcessList.NATIVE_ADJ : -1000
2 System ProcessList.SYSTEM_ADJ : -900
3 Persistent ProcessList.PERSISTENT_PROC_ADJ : -800
4 Persistent Service ProcessList.PERSISTENT_SERVICE_ADJ : -700
5 Foreground ProcessList.FOREGROUND_APP_ADJ : 0
6 Visible ProcessList.VISIBLE_APP_ADJ : 100
7 Perceptible ProcessList.PERCEPTIBLE_APP_ADJ : 200
8 Backup ProcessList.BACKUP_APP_ADJ : 300
9 Heavy Weight ProcessList.HEAVY_WEIGHT_APP_ADJ : 400
10 A Services ProcessList.SERVICE_ADJ : 500
11 Home ProcessList.HOME_APP_ADJ : 600
12 Previous ProcessList.PREVIOUS_APP_ADJ : 700
13 B Services ProcessList.SERVICE_B_ADJ : 800
14 Cached ProcessList.CACHED_APP_MIN_ADJ : 900
文章作者: Li Shuaiqi
文章链接: https://lishuaiqi.top/2016/06/13/Process6-dumpsysMeminfoToShowAdj/
版权声明: 本博客所有文章除特别声明外，均采用 CC BY-NC-SA 4.0 许可协议。转载请注明来自 Li Shuaiqi's Blog！
Process进程